This document summarizes a presentation on developing robotic systems for upper limb neurorehabilitation. It discusses developing human-robot symbiosis through wearable robots that allow natural control and safe interaction. The presentation covers robotic hands that can perform actions like reaching, grasping, and manipulating objects, and how these can be used for motor therapy, prosthetics, and personal assistance. Developing systems that provide body ownership during physical human-robot interaction is also discussed.

1. Sviluppo
di
sistemi
robo0ci
per
la
neuroriabilitazione
dell'arto
superiore
Maria
Chiara
Carrozza
Scuola
Superiore
Sant’Anna
Pisa,
Italy
XII
CONGRESSO
SIAMOC
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
1
Bosisio
Parini,
28
se@embre-­‐1
o@obre
2011

2. Content
• Symbiosis
• AcDons
&
FuncDons
• Hand
and
Brain
• RehabilitaDon
and
Assistance
• Future
perspecDves
(and
lower
limb)
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
2

3. Content
• Symbiosis
• AcDons
&
FuncDons
• Hand
and
Brain
• RehabilitaDon
and
Assistance
• Future
perspecDves
(and
lower
limb)
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
3

4. Human-­‐robot
symbiosis
Is
“physical”
human-­‐exoskeleton
symbiosis
doable?
In
1960s,
in
Man-­‐Computer
symbiosis
,
J.C.R.
Licklider
formulated
a
vision
of
human-­‐computer
symbiosis
in
which
computers
and
humans
would
become
fluidly
interdependent
and
share
goals
In
2010s,
in
many
tasks,
human
and
computer
share
goals
and
are
interdependent
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
4

5. Symbiosis:
requirements
• Wearability
• Natural
control
• Safe
interacDon
• Body
ownership
• …..
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
5

6. Human-­‐robot
symbiosis
• Moving
from
human-­‐computer
to
physical
human-­‐robot
(or
human-­‐exoskeleton)
symbiosis
requires
addressing
some
design
issues:
Wearability
Natural
Control
Safe
Interac=on
safe
and
comfortable
Non-­‐invasive
user
mo0on
Distributed
sensoriza0on
physical
human-­‐robot
inten0on
detec0on
and
gentle
of
the
physical
human-­‐
interac0on
assistance
robot
interface
NEUROExos
elbow
exoskeleton
Adap0ve
oscillators
for
assis0ng
Sensorized
cuff
for
lower-­‐limb
rhythmic
mo0ons
exoskeletons
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
6

7. AcDons
and
FuncDons
• AcDons:
• Reaching,
Touching,
Grasping,
Feeling,
ManipulaDng
• FuncDons:
– LocomoDon,
navigaDon,
manipulaDon
– AcDvity
of
daily
living
– Therapy:
motor
recovery
– Replacement:
prostheDcs
– Enhancement:
personal
assistance
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
7

8. Content
• Symbiosis
• AcDons
&
FuncDons
• Hand
and
Brain
• RehabilitaDon
and
Assistance
• Future
perspecDves
(and
lower
limb)
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
8

9. Reaching,
Grasping,
Touching,
ManipulaDng
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
9

10. Touching
and
Grasping
B.
B.
Edin,
L.
Ascari,
L.
Beccai,
S.
Roccella,
J.-­‐J.
Cabibihan,
and
M.C.
Carrozza,
“Bio-­‐inspired
sensorizaDon
of
a
biomechatronic
robot
handfor
the
grasp-­‐and-­‐lic
tasks,”
Brain
Res.
Bull.,
vol.
75,
pp.
785–795,
2008
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
10

11. Wearability
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
11

12. Body
ownership
In
the
"rubber-­‐hand
illusion,"
the
sight
of
brushing
of
a
rubber
hand
at
the
same
Dme
as
brushing
of
the
person's
own
hidden
hand
is
sufficient
to
produce
a
feeling
of
ownership
of
the
fake
hand.
We
shown
previously
that
this
illusion
is
associated
with
acDvity
in
the
mulDsensory
areas,
most
notably
the
ventral
premotor
cortex
(Ehrsson
et
al.,
2004Go).
However,
it
remains
to
be
demonstrated
that
this
illusion
does
not
simply
reflect
the
dominant
role
of
vision
and
that
the
premotor
acDvity
does
not
reflect
a
visual
representaDon
of
an
object
near
the
hand.
To
address
these
issues,
we
introduce
a
somaDc
rubber-­‐
hand
illusion.
The
experimenter
moved
the
blindfolded
parDcipant's
lec
index
finger
so
that
it
touched
……
h@p://www.dailymoDon.com/video/xhb0kd_roboDc-­‐hand-­‐that-­‐feels-­‐real_tech
H.
H.
Ehrsson,
N.
P.
Holmes,
and
R.
E.
Passingham,,”
Touching
a
Rubber
Hand:
Feeling
of
Body
Ownership
Is
Associated
with
Ac0vity
in
Mul0sensory
Brain
Areas”,
The
Journal
of
Neuroscience,
November
9,
2005,
25(45):10564-­‐10573
M.
L.
Blefari,
C.
Cipriani,
M.
C.
Carrozza,
"A
Novel
Method
for
Assessing
Sense
of
Body
Ownership
Using
Electroencephalography,"
IEEE
Trans.
Biomedical
Engineering,
2010,
p.
1,
99,
10.1109/TBME.2010.2076282
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
12

13. Three
main
challenges:
(1)
(2)
RoboDc
hand
able
to
Suitable
interface
for
perform
dexterous
acDons
controlling
percepDon
and
and
sense
objects
acDon
Interface
bi-­‐direcDonal
(3)
Sense
of
Body
Ownership
M.
C.
Carrozza,
G.
Cappiello,
S.
Micera,
B.
B.
Edin,
L.
Beccai,
and
C.
Cipriani,
“Design
of
a
cyberneDc
hand
for
percepDon
and
acDon,”
Biol.
Cybern.,
vol.
95,
no.
6,
pp.
629–644,
2006
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
13

14. Content
• Symbiosis
• AcDons
&
FuncDons
• Hand
and
Brain
• RehabilitaDon
and
Assistance
• Future
perspecDves
(and
lower
limb)
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
14

15. Why
a
roboDc
model
of
the
human
arm
• NEURARM
is
intended
to
test
hypotheses
of
the
human
arm
motor
control
system
• A
model
that
is
“under
full
control
of
the
experimenter”
• A
roboDc
model
is
a
powerful
tool
to
overcome
the
possible
“reality
gap”
of
mathemaDcal
models
• A
roboDc
model
can
complement
results
from
either
numerical
simulaDons
or
measurements
on
human
arm
(e.g.
Burdet
et
al.,
2001)
Lenzi
T,
ViDello
N,
McIntyre
J,
Roccella
S,
Carrozza
MC,
(2010),
A
roboDc
model
to
invesDgate
the
human
motor
control,
Biological
Cyberne=cs,
in
press,
2011.
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
15

16. Design
goals
for
the
NEURARM
Mimic
the
dynamics
and
control
structure
of
the
human
arm
• ac0vely-­‐adjustable
passive
compliance
 To
invesDgate
neuroscien0fic
– tendon
driven
hypotheses
on
motor
control
and
– antagonisDc
muscle
pairs
test
possible
benefits
for
controlling
– nonlinear
spring-­‐like
actuators
roboDc
arDfacts
• similar
limb
kinema0cs
and
dynamics
 ApplicaDon
field:
robo0cs
for
– limb
inerDa
rehabilita0on
– joint
sDffness
ranges
RoboDc
Biological
model
CyberneDc
model
Bio-­‐inspired
roboDc
arDfact
implementaDon
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
16

17. Remote
and
antagonist
actuaDon
• Two
remote
and
antagonist
actuaDon
units
• Each
unit
is
a
series
of
a
non-­‐linear
spring,
a
contracDle
element
and
a
steel
wire
rope
• Two
control
strategies:
– passive
compliance
control,
for
robot-­‐in-­‐charge
therapy
– torque
control,
for
pa=ent-­‐in-­‐charge
therapy
 Despite its greater
complexity, pros are:
 compared to gearhead
DC motors, a low
output impedance
over all the frequency
spectrum
 compared to series
elastic actuators, an
adjustable hardware
compliance
N.
ViDello,
T.
Lenzi,
S.M.M.
De
Rossi,
S.
Roccella,
M.C.
Carrozza,
“A
sensorless
torque
control
for
AntagonisDc
Driven
Compliant
joints”,
Mechatronics,
vol.
20(3),
pp.
355-­‐367,
2010.
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
17

18. Adjustable
Rest-­‐Length
and
Non-­‐linear
Springs
Reciprocally
shiing
the
length-­‐tension
curve
of
the
opposing
springs
causes
a
shic
of
Equilibrium
Point.
Concomitantly
shiing
the
length-­‐tension
curves
of
the
opposing
springs
increases
the
sDffness
in
each
muscle
for
the
same
Equilibrium
Point.
 Changing
the
Equilibrium
Point
is
achieved
by
reciprocal
acDvaDon
 Net
sDffness
may
be
increased
by
co-­‐acDvaDon
∴
Nonlinear
springs
make
EPH
control
viable
and
similar
to
biological
muscles
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
18

19. Actuator
Control
The
Equilibrium
Point
Hypothesis
based
control
Antagonist Antagonist
pistons move in pistons move in
the same opposite
direction direction
STIFFNESS POSITION
regulation regulation
XExt/Flx = XCOM ± XDIFF
XCOM – common mode displacement  joint stiffness
XDIFF – differential mode displacement  joint equilibrium position
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
19

20. 4-­‐DOF
passive
mechanism
• From
biomechanics*:
– human
elbow
is
a
LOOSE
hinge
joint
– because
of
the
intra-­‐
and
–inter
subject
variability,
elbow
axis:
– traces
a
double
conic
frustum
(max
values
for
βh
and
βf
are:
10°and
6)
along
the
flexion-­‐extension
movement
– forms
an
angle
with
AH
of
80°-­‐92°
– forms
an
angle
with
AML
of
±5°
• We
need
a
passive
mechanism
which
compensates
for:
– the
rotaDons
of
elbow
flexion
axis
in
both
frontal
and
horizontal
planes
– compensate
for
the
human-­‐robot
joint
axes
misalignment
(Adapted
from
I.A.
KAPANDJI
Physiology
of
Human
Joints
)
*Bo@lang
et
al.,
1998;
Bo@lang
et
al.,
2000;
Duck
et
al.,
2003
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
20

21. 4-­‐DOF
passive
mechanism
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
21

22. Passive
compliance
control
• Passive
compliance
control:
– inspired
by
the
Equilibrium
Point
Hypothesis
– independent
control
of
joint
equilibrium
posiDon
and
sDffness,
by
means
of
the
independent
control
of
the
pistons
posiDons
[Nm]
[deg]
StaDc
characterizaDon
of
the
tunable
Step
response
of
the
posiDon
control
for
passive
compliance;
sDffness
range:
three
different
values
of
sDffness
28-­‐72
Nm/rad
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
22

23. Torque
control
• Two
independent
closed-­‐loop
force
controllers
– (a)
step
response
of
the
force
control
• raising
Dme:
0.54
s
– (b)
closed-­‐loop
Bode
diagram:
• -­‐3
dB
bandwidth:
≈10
Hz
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
23
(a)
(b)

24. AdapDve
oscillators
for
moDon
assistance
Experiment
with
NEUROExos
(in
torque
control
modality),
conducted
at
ARTS
Lab,
Scuola
Superiore
Human-­‐robot
cross-­‐adapta0on
Sant’Anna
in
collaboraDon
with
EPFL
(Switzerland)
through
synchroniza0on
R.
Ronsse,
N.
ViDello,
T.
Lenzi,
J.
van
den
Kieboom,
M.C.
Carrozza,
A.J.
Ijspeert,
“Human-­‐robot
synchrony:
flexible
assistance
using
adapDve
oscillator”,
IEEE
Transac=ons
on
Biomedical
Engineering,
vol.
58(4):1001-­‐12,
2011.
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
24

25. RehabilitaDon
and
Assistance
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
25

26. Materials
and
methods:
the
assisDve
strategy
• Block
diagram
of
the
assisDve
strategy
• The
adapDve
oscillator
is
a
modified
Hopf
oscillator
that
can
learn
the
frequency
ω(t),
amplitude
α1(t)
and
offset
α0(t):
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
26

27. Results:
adapDve
oscillator
tracks
the
actual
joint
angle
AdapDve
oscillator
tracks
elbow
joint
angle
with
no
delay
True
signal
in
black,
oscillator-­‐based
esDmate
in
dashed
gray

28. 1st
case
(frequency
constant):
EMG
profile
ReducDon
of
EMG
acDvity,
for
both
biceps
and
triceps,
is
more
evident
if
we
look
at
the
single-­‐cycle
EMG
profile
averaged
over
all
subjects
and
cycles
no-­‐exoskeleton
no-­‐assistance
(k=0)
low-­‐
assistance
(k=0.33)
high
assistance
(k=0.5)
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
28

29. Assistance
through
Powered
• The
big
challenge
Exoskeletons
– understand
the
intenDon
of
the
user
and
react
appropriately
to
provide
the
required
assistance
in
Dme
• The
know
strategy
1. Es0mate
with
the
best
possible
accuracy
the
torque
needed
to
perform
the
desired
movement
2. Provide
a
constant
fracDon
of
that
torque
to
the
user
• The
open
quesDons
– Is
accuracy
the
“holy
grail”
of
robot-­‐mediated
assistance?
– Is
the
“constant
fracDon”
paradigm
strictly
needed
for
assisDng
effecDvely
humans?
– How
to
take
into
account
the
user’s
reacDon?
• What
does
it
mean
assistance
from
a
Motor
Learning
perspecDve?
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
29

30. The
ProporDonal
EMG
controller
• Our
hypothesis
– motor
learning
can
compensate
for
torque
esDmate
imprecision,
without
adding
further
cogniDve
effort
to
the
user
• Our
goal
– Verify
the
closed-­‐loop
usability
of
a
simplified
assisDve
controller
• proporDonal
EMG
control
• preferred
gain
directly
chosen
by
subjects
The
NEUROExos
elbow
exoskeleton
Block
diagram
of
the
simplified
assisDve
controller
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
30

31. Experimental
Procedure
(1/2)
1. PreparaDon
(electrodes
placement)
– sEMG
from
the
biceps
brachii
and
triceps
brachii
• pre-­‐gelled
Ag/AgCl
(Pirronse&Co.,
Italy)
• digiDzed
@1Khz
with
internal
band-­‐pass
filter
(10-­‐500Hz)
(Noraxon)
2. Preferred
gain
selecDon
– Subjects
don
NEUROExos
and
perform
unconstrained
flexion-­‐extension
movement
– Subjects
increase
the
gains
gradually
using
a
knob
as
long
as
they
feel
comfortable
with
assistance
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
31

32. Experimental
Procedure
(2/2)
3. Behavioral
experiments
– Rhythmic
flexion/extension
movement
against
gravity
at
a
target
pace
and
amplitude
supplied
through:
• augmented
visual
feedback
• acousDc
cueing
– Variable
assistance
level
•
50%-­‐100%-­‐150%
of
the
preferred
gain
previously
chosen
– Different
movement
condiDon
• hand-­‐free
movement
(trial
1)
• dumb-­‐bell
licing
(trial
2)
• variable
target
pace
(trial
3)
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
32

33. Results
(1/2)
• PosiDon
velocity
and
acceleraDon
profiles
are
not
altered
by
the
assistance
• LE
decrease
with
the
assistance
indicaDng
an
effecDve
effort
reducDon
Steady-­‐state
profiles
of
posiDon
(A),
velocity
(B)
,
acceleraDon
(C)
and
biceps
linear
envelope
(D)
for
one
representaDve
subject.
Profiles
were
obtained
by
resampling
the
actual
trajectories
over
1000
equally
spaced
samples
for
each
cycle,
then
averaging
on
the
steady-­‐state
cycles
of
each
gain
separately.
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
33

34. Discussion
• Subjects
could
keep
the
full
control
of
their
arm
movement
despite
the
acDon
of
the
EMG-­‐proporDonal
assisDve
controller.
– with
the
EMG
assistance
on,
parDcipants
successfully
performed
the
rhythmic
task
independently
of
the
presence
of
an
addiDonal
weight
and
the
required
movement
pace
• In
all
tested
condiDons
parDcipants
could
reduce
the
effort
spent
for
generaDng
the
arm
movement
as
shown
by
the
considerable
biceps
IEMGs
drop
off
• Very
fast
adapta0on
of
subjects
to
the
disturbances
induced
by
the
EMG
controller.
– all
parDcipants
could
recover
the
target
movement
pace
and
amplitude
within
3
cycles
acer
an
abrupt
assistance
acDvaDon,
regardless
of
the
specific
movement
condiDon
experienced
during
the
trials.
– IEMG
recordings
reached
a
staDonary
level
within
the
same
number
of
cycles
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
34

35. Design
goals
for
Handexos
• Development
of
an
exoskeleton
for
rehabilitaDon
and
funcDonal
recovery
of
the
human
hand
• System
requirements
– Matching
the
kinemaDc
requirements
of
the
human
hand
• Inter-­‐subjects
anthropometric
variability
– Comfortable
physical
human-­‐robot
interacDon
– Minimum
encumbrance
– Low
inerDa
of
the
moving
parts
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
35

36. HANDEXOS:
1st
prototype
 Shell-­‐like
structures
for
the
link
 MinimizaDon
of
the
overall
size
 Comfortable
distributed
physical
human-­‐robot
interface
 Proximal
slider
crank-­‐like
mechanism
to
transmit
the
driving
torque
onto
the
human
MCP
joint
while
minimizing
the
undesired
forces
loading
the
MCP
arDculaDon
 Underactuated
remote
actuaDon
 Free
palm
area
 Chiri et al., IROS, 2009
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
 Chiri ert al., IEEE Transactions on Mechatronics, 2011
36
 S. Roccella et al., Wearable Mechatronic Device, Pub. No.: WO/2009/016478

37. Index
finger
module
Finger
mechanism,
exploded
view
 3
acDve
rotaDonal
joints
for
finger
flexion/
extension
 1
passive
rotaDonal
joint
for
finger
abducDon
adducDon
 1
passive
translaDonal
joint
for
the
kinemaDc
coupling
of
the
human/exoskeleton
MCCP
aCarrozza
13-­‐10-­‐2011
©
Maria
hiara
xes
37

38. ActuaDon
System
 Light
weight
 Modular

Simply
reconfigurable
 Remotely
located
with
respect
to
the
hand
 It
is
composed
of
an
actuated
extension
unit
(top
blue
box)
and
a
passive
flexion
module
(down
red
box)
 The
transmission
system
is
based
on
steel
wire
ropes
routed
through
spiral-­‐
spring
Bowden
cables
 external
diameter
1.6
mm
 internal
diameter
0.8
mm
 The
tendon
cable
is
pulled
by
a
linear
slider
driven
by
a
dc
motor
(18.10
W)
through
a
planetary
gear
with
a
reducDon
raDo
of
43
 The
moDon
is
transmi@ed
to
a
lead
screw
(pitch
0.5
mm)
by
means
of
spur
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
38
gears

39. Sensors
 Lower
shells
are
equipped
with
three
force
sensors
to
sense
the
interacDon
force
onto
the
palmar
side
 The
extension
unit
is
equipped
with
two
Hall
sensors
which
serve
as
limit
switches
 Cable
force
is
sensed
through
strain
gauges
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
39

40. Control
System
 The
low-­‐level
layer
running
at
5
kHz
on
a
standalone
moDon
controller
implements
the
posiDon
control
of
the
linear
slider
 The
high-­‐level
layer
running
on
a
remote
PC
at
100
Hz
sets
the
desired
posiDon
of
the
linear
slider
in
accordance
with
the
rehabilitaDve
task
to
be
performed,
and
it
monitors
the
cable
force
for
reverDng
the
slider
moDon
in
the
case
of
force
overload
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
40

41. (11) Experimental
results
Comparison
of
finger
kinema0cs
with
and
without
wearing
HANDEXOS
 All
subjects
could
easily
wear
HANDEXOS
and
perform
extension/flexion
tasks
 The
level
of
similarity
of
joint
trajectories
between
NE
and
NA
modes
was
assessed
by
calculaDng
the
Pearson
product
moment
correlaDon
 Thanks
to
its
passive
DOFs,
HANDEXOS
could
fit
the
hand
anthropometry
of
each
subject,
and
none
of
them
reported
any
discomfort
in
wearing
the
device
Human
MCP,
PIP
and
DIP
trajectories
acquired
during
NA
(no
acDon,
dashed
line)
and
NE
(no
exoskeleton,
solid
line)
modes
for
subject
#1.
Wearing
HANDEXOS
does
not
modify
the
finger
extension/flexion
mo0on
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
41

42. Experimental
results
Evalua0on
of
the
undesired
transla0onal
force
ac0ng
on
the
human
MCP
ar0cula0on
 Small
undesired
reciprocal
translaDon
exists
between
the
device
and
the
user’s
hand
during
flexion-­‐extension
moDon
tasks
 The
maximum
esDmated
translaDons
force
acDng
on
the
human
MCP
Acquired
(θ1
in
solid
black
line)
and
esDmated
(θ’1
in
dashed
arDculaDon
in
a
task
of
assisted
black
line)
HANDEXOS
MCP
trajectories
for
subject
#1.
The
extension
was
about
8
N
grey
line
represents
the
human
MCP
trajectory
(q1).
AVERAGE
AND
STANDARD
DEVIATION
OF
RMS
AND
MAX
ERR
FOR
EACH
SUBJECT
HANDEXOS
does
not
overload
MCP
aMaria
Chiara
Carrozza
than
the
ac0vi0es
of
daily
living
13-­‐10-­‐2011
©
r0cula0on
more
42

43. Content
• Symbiosis
• AcDons
&
FuncDons
• Hand
and
Brain
• RehabilitaDon
and
Assistance
• Future
perspecDves
(and
lower
limb)
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
43

44. Work
on
lower
limb:
measurement
of
HRI
force
• The
need
for
measuring
interacDon
– How
the
machine
is
dynamically
interacDng
with
the
user
• How
interacDon
is
transmi@ed
– Cuffs
– Orthoses
• How
interacDon
is
measured
HAL-­‐5
MIT-­‐leg
exos
Lokomat
– InteracDve
torque
• Model
–
based
approaches
– InteracDve
force
• Load
cells
(Lokomat)
• Spring-­‐based
force
sensors
(MIT
Exoskeleton)
• Strain
gauges
(HAL)
Michingan
AFO
MIT
AFO
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
44

45. Sensorized
cuff
Shielded
cable
Rigid
plasDc
frame
Connector
Flexible
belt
Skilsens
pad
• Soc
tacDle
sensor
does
not
affect
the
comfort
• The
posiDon
and
number
of
sensors
can
be
changed
• Adaptable
to
all
cuff
sizes
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
45

46. The
Skilsens
tacDle
sensing
technology
• Skilsens
technology
• Each
sensor
can
be
adapted:
– Size
and
shape
– Force
range
– SpaDal
resoluDon
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
46

47. Results:
walking
experiments
• This
sensory
apparatus
can
discriminate
between
two
walking
condiDons:
– Zero-­‐torque
control
– Viscous
field:
10
Nm/rad·∙s-­‐1
Increased
local
30
kPa
peaks
pressure
S.M.M.
De
Rossi,
N.
ViDello,
T.
Lenzi,
R.
Ronsse,
B.
Koopman,
A.
Persiche„,
F.
Vecchi,
A.J.
Ijspeert,
H.
van
der
Kooij,
M.C.
Carrozza,
“Sensing
pressure
distribuDon
on
a
Lower-­‐Limb
Exoskeleton
Physical
Human-­‐Machine
Interface”,
Sensors,
vol.
11(1),
pp.
207-­‐227,
2010.
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
47

48. Summary
of
features
• New
Prototype
Pros:
• Maximum
wearability
and
comfort
• Allows
Long-­‐term
recordings
in
a
normal
shoe
• Coverage
of
most
of
the
plantar
area
• Allows
to
compute
relaDve
posiDon
of
CoP
and
total
GRF
• Good
single-­‐cell
esDmaDon
accuracy
• High
temporal
resoluDon
for
high-­‐
and
low-­‐level
data
• Possible
improvements:
• Decrease
sensor
size
• Reduce
size
of
onboard
electronics
• Integrate
electronics
on
the
sensor
board
• On-­‐board
data
logging
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
48

49. Acknowledgments
• NANOBIOTACT:
Nanoengineering
BiomimeDc
TacDle
Sensors
(Project
EU-­‐FP6-­‐NMP-­‐033287)
• NANOBIOTOUCH:
Nano-­‐resolved
mulD-­‐scale
invesDgaDons
of
human
tacDle
sensaDons
and
Dssue
engineered
nanobiosensors
(Project
EU-­‐FP7-­‐
NMP-­‐228844)
• WAY
Wearable
interfaces
hAnd
funcDon
recoverY
FP7-­‐ICT-­‐2011-­‐
288551
• The
EVRYON
CollaboraDve
STREP
Project:
“EVolving
MoRphologies
for
Human-­‐Robot
SYmbioDc
InteracDON”,
FP7-­‐ICT-­‐2007-­‐3-­‐231451
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
49

50. NeuroroboDcs
@
The
BioroboDcs
InsDtute
ChrisDan
Cipriani
(Assistant
Professor)
Fabrizio
Vecchi
(Research
manager)
Stefano
Roccella
(Senior
research
assistant)
Nicola
ViDello
(Post-­‐Doc)
Alessandro
Persiche„
(Post-­‐Doc)
Michela
Aquilano
(Post-­‐Doc)
Filippo
Cavallo
(Post-­‐Doc)
Calogero
Oddo
(Post-­‐Doc)
Marco
Controzzi
(PhD
student)
Manuele
Bonaccorsi
(PhD
Student)
Tommaso
Lenzi
(PhD
student)
Stefano
De
Rossi
(PhD
student)
Azzurra
Chiri
(PhD
student)
Marco
Cempini
(PhD
Student)
Maria
Laura
Blefari
(PhD
Student)
Gunter
Kanitz
(PhD
Student)
Marco
D’Alonzo
(PhD
Student)
Francesco
Giovacchini
(Research
assistant)
Marco
DonaD
(Research
assistant)
13-­‐10-­‐2011
©
Maria
Chiara
Carrozza
50